AdaHuman generates high-fidelity, reposable 3D avatars from a single image by adding pose conditioning and local body-part refinement to multiview diffusion with Gaussian splat reconstruction.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AdaHuman: Animatable Detailed 3D Human Generation with Compositional Multiview Diffusion
AdaHuman generates high-fidelity, reposable 3D avatars from a single image by adding pose conditioning and local body-part refinement to multiview diffusion with Gaussian splat reconstruction.